Credible Threats

Have you ever heard someone say, “If you do that, I’ll never help you again”? students, in game theory, that kind of statement is only powerful if it is believable. If the person making the threat would not actually follow through when the time comes, then the threat is not credible. In dynamic games, this matters a lot because players often make promises and threats before later decisions are reached. The main goal of this lesson is to help you tell the difference between threats that shape behavior and threats that are just empty words.

What Makes a Threat Credible? 🤔

A credible threat is a plan of action that a player would really want to carry out if the situation arrives. In other words, it is not enough to say, “I will punish you.” The key question is whether punishment is still the best choice after the other player has moved.

This idea is easiest to see in a sequential game. Suppose Player 1 chooses whether to enter a market, and Player 2 can respond by starting a price war. Player 2 may threaten to cut prices so low that Player 1 loses money. But if Player 1 actually enters, and a price war would also hurt Player 2 even more, then Player 2 would not want to carry out the threat. In that case, the threat is not credible.

Credibility depends on incentives at the moment a decision is made. A threat is credible only if the action is optimal when the relevant decision node is reached. This is one reason backward induction is so important in dynamic games. By reasoning from the end of the game backward, we can check whether later actions really make sense.

A helpful way to think about it is like a sports coach saying, “If you miss practice, you will be benched.” If the coach always benches players who miss practice, the threat is believable. If the coach never actually does it, players will stop taking the warning seriously. In game theory, believable threats matter because rational players anticipate what will really happen, not just what was said earlier ⚽.

Non-Credible Off-Path Behavior

A threat is often called non-credible when it describes what a player says they will do after an event that does not occur in the equilibrium path. This is called off-path behavior. The “path” is the sequence of actions actually taken in equilibrium. If some unexpected move happens, the game may reach a different decision point, and the earlier threat may be tested.

The important thing is that off-path behavior must still be rational. A player cannot rely on an action that would clearly be worse for them just to scare the other player. If a threat is only useful as a warning but not as a real plan, it is not believable.

For example, imagine two companies competing. Company A considers entering a new city. Company B threatens to open a temporary store right next door and sell at a loss to drive A out. If B would lose a lot of money by actually doing this, then the threat may be empty. A rational entrant knows B is unlikely to carry it out, so the threat does not strongly affect A’s decision.

This shows why not every equilibrium that contains a threat is convincing. Some strategies may specify actions that sound severe, but once the moment arrives, the player would rather choose something else. In subgame perfection, we rule out these kinds of plans because they depend on future behavior that is not optimal when the future becomes present.

A simple rule students can remember is this: if the action is only good for scaring the other player but bad to actually do later, it is probably non-credible.

Why Credibility Matters in Equilibrium

Equilibrium in game theory means that every player’s strategy is a best response to the others’ strategies. But in dynamic games, that is not always enough. A strategy profile can be a Nash equilibrium even if it uses threats that would never be carried out. That is why we need stronger ideas like subgame perfect equilibrium.

Subgame perfection requires that strategies be optimal not just at the start of the game, but also in every subgame. A subgame is any part of the game that starts at a decision point and contains everything that follows. This means players must have a rational plan for every possible future situation, including those that may be unlikely.

Credibility is central here because a threat only influences behavior if others believe it. Suppose Player 1 is deciding whether to take a small profit now or continue in a repeated interaction. If Player 2 threatens to retaliate forever after any small deviation, Player 1 might obey only if Player 2 would truly retaliate. But if retaliation is costly and would not be chosen later, the threat is not credible, and Player 1 should ignore it.

This is why credible threats help separate real strategic power from empty intimidation. In economics, business, politics, and negotiations, people often try to influence others with promises and warnings. Game theory teaches us to ask a sharper question: “Would the player still want to do this when the moment comes?” If the answer is no, the threat cannot support a believable equilibrium.

A Step-by-Step Example

Let’s build a very simple sequential game to see how credibility works. Suppose Player 1 can choose $E$ for enter or $N$ for not enter. If Player 1 chooses $N$, both players get a payoff of $2$ and $2$. If Player 1 chooses $E$, then Player 2 can choose $F$ for fight or $A$ for accommodate.

The payoffs are:

If Player 2 chooses $F$, payoffs are $(-1,-1)$.
If Player 2 chooses $A$, payoffs are $(3,1)$.

Now ask: what would Player 2 actually do if Player 1 entered? Comparing Player 2’s payoffs, $1$ from accommodating is better than $-1$ from fighting. So Player 2 would choose $A$, not $F$.

That means a threat like “If you enter, I will fight” is not credible, because fighting is not Player 2’s best action after entry. Since Player 1 knows this, Player 1 expects accommodation and may choose $E$.

Here is the logic backward:

If entry happens, Player 2 chooses $A$.
Knowing that, Player 1 compares the payoff from entering, which is $3$, with the payoff from not entering, which is $2$.
So Player 1 chooses $E$.

The resulting equilibrium does not rely on a fake punishment. If someone claimed to threaten fighting, that claim would be empty because Player 2 would never want to follow through. This is the basic idea of non-credible threats 👀.

Backward Induction and Credibility

Backward induction is the method of solving a game from the end toward the beginning. It works especially well in finite dynamic games because it reveals what each player will actually choose at every decision point. Since credible threats must be optimal when the time comes, backward induction naturally filters out threats that cannot be trusted.

This is why subgame perfect equilibrium and credible threats go together. A subgame perfect equilibrium is an equilibrium where the strategy is a Nash equilibrium in every subgame. If a threat requires a player to choose an action that is bad for them in a later subgame, then that strategy is not subgame perfect.

Think of it like planning a road trip. If students says, “If traffic is bad, I will take the longer route,” but the longer route is obviously much worse at that time, then the plan is not realistic. A good plan must still make sense when the situation actually arrives. Game theory demands the same kind of realism.

This does not mean players never make threats. It means only some threats matter in equilibrium analysis. Believable threats can shape expectations and behavior. Non-credible threats are ignored because rational opponents know they will be abandoned when the time comes.

How to Spot a Credible Threat

When analyzing a game, students can use a few questions to test credibility:

Would the threatening player actually prefer to carry out the threat if the relevant situation occurred?
Is the threat part of a later decision after some move by the other player?
Does the action remain a best response in that subgame?
Would backward induction predict the same action?

If the answer to the first and third questions is yes, the threat is credible. If not, it is non-credible.

A threat can also be credible even if it is harmful to both players, as long as it is still better for the threatening player than the alternatives at that moment. For example, in some bargaining situations, a player may credibly reject a bad offer because accepting would be even worse. Credibility is about incentives, not just about whether an action seems harsh.

The main takeaway is that credibility is not about loudness, anger, or confidence. It is about whether the threatened action fits the player’s incentives once the game reaches that point.

Conclusion

Credible threats are actions a player would actually carry out if the relevant moment arrives. Non-credible threats are promises of punishment that would be abandoned later because they are not in the player’s interest. In dynamic games, credibility matters because rational players anticipate future choices, not just words. That is why subgame perfection removes equilibria that depend on empty threats.

If you remember one idea from this lesson, students, let it be this: a threat influences behavior only when it is believable, and a threat is believable only when the player would rationally follow through.

Study Notes

A credible threat is a threat a player would actually carry out if the situation arose.
A non-credible threat is a threat that is not optimal when the time comes.
Off-path behavior is what players do after an unexpected move that is not part of the equilibrium path.
Subgame perfect equilibrium removes strategies that depend on non-credible threats.
Backward induction helps test whether later actions are believable.
A threat is credible only if it is a best response in the relevant subgame.
Rational players do not believe empty punishment if the threatening player would later prefer a different action.
Credibility is important in business competition, bargaining, politics, and everyday strategic interactions.